22.01 Reinstall goes wrong

KOM

I have an SG-1100 happily running 21.05.2. I wanted the resiliency of ZFS so I grabbed the 22.01 install image and tried to install fresh. Note that I have successfully done this in the past with earlier versions.

The run usbrecovery process seemed to run well and finish correctly. After reboot, I get this:

Card did not respond to voltage select!
mmc_init -95, time 41
** Bad device mmc 0 **
...
...
ERROR: Cannot open /boot/lua/loader.lua: invalid argument.

I know that everything works until it breaks, but this router was chugging along just fine until the moment I tried to reinstall. I'm having a hard time understanding how the MMC managed to die at this very moment.

In the meantime I dug out my old Mikrotik HeX-PoE and panicked to remember how to access and configure it while my family yelled at me. Not how I expected to spend an hour.

Any suggestions would be appreciated before I bug Netgate Support when I know they're likely being slammed.

stephenw10

What image did you use exactly?

That error is what you get with the older uboot version trying to boot the newer installer image. You need the compatibility 'image' the first time you boot it which will update uboot.

Steve

KOM

@stephenw10 I was sent a link to:

pfSense-plus-compat-recovery-22.01-RELEASE-aarch64.img.gz

stephenw10

Hmm, that should be the correct image. I would open a ticket to diagnose it further then.

Steve

KOM

@stephenw10 Damn. I was hoping you had a magic bullet.

stephenw10

The error is exactly what you see if you don't use that image. If you try to boot the 2100 recovery image for example. I would double check it's definitely that image and that the file hash matches. But otherwise the full console log should help the support guys.

Steve

KOM

@stephenw10 I opened a ticket and went back & forth with them. They think it might be bad mmc, usb stick or usb port but I've tried two different sticks in both USB ports. I asked if there is a way to read/write test the mmc to rule it out and am waiting for a reply.

stephenw10

Ok, I think the issue here is that the installer for some reason did not update uboot.

What was unclear last night was where this error was happening. I incorrectly assumed it was when you were trying to boot the USB stick which would only be the wrong image as I said bit was this in fact after appearing to install successfully and then trying to boot the result from eMMC?

If that's the case it's because uboot did not update for some reason and the older uboot cannot boot from ZFS.
Check the uboot version show at boot. If it's still the older version it will show as:

U-Boot 2017.03-armada-17.10.2-g6a6581a-dirty (Nov 01 2018 - 16:04:57 -0300)

If that is the case you should still be able to install as UFS and boot that.

Steve

KOM

@stephenw10 Yes it still shows the older version of Uboot, and as you guessed reinstalling with UFS works fine. I've reported back to Netgate and will update when I hear something from them.

stephenw10

Does it show that exact uboot version or one from Nov 22nd 2018? There might be a clue here...

KOM

@stephenw10

U-Boot 2017.03-armada-17.10.2-g6a6581a-dirty (Nov 01 2018 - 16:04:57 -0300)

stephenw10

Ok, that might be the issue. If you can please run the following command:

[22.01-RELEASE][admin@1100-3.stevew.lan]/root: sh -c 'dd if=/dev/flash/spi0 bs=64k count=1 2>/dev/null | sha256'
6eeae92008003686d912066830858e4af0fa8476b23a719c10f60e1edfb8cfeb

The updater checks for known sha256 sums and it appears yours may not be expected. That is the updated uboot shown on my 1100 here.

One you have booted into 22.01 using UFS though you should be able to manually update uboot and then re-install ZFS if that turns out to be the issue.

Steve

KOM

@stephenw10 My output is 1b6998c83988bb0e53500606d340d119f07846e4058afb8eb55c3eca07ffbcef

How does one manually update U-Boot?

stephenw10

If you repeat that command does it come back the same every time?

KOM

@stephenw10 Yes

stephenw10

Confirming this locally....

KOM

@stephenw10 Should I leave my unit untouched in case you need to refer to something or is it ok for me to proceed getting it working with UFS? I rely on this for VPN to work access and it needs to be up & ready by Thursday evening.

rcoleman-netgate

@kom I would just get it back online for your VPN. We're working on tracking down a unit that has this same UBoot internally.

stevie_0815

@rcoleman-netgate

Hi,

I have a recently purchased SG-1100 and requested the new firmware so that i could start working with a ZFS-based install. I'm getting the same error on my device:

U-Boot 2017.03-armada-17.10.2-g6a6581a-dirty (Nov 01 2018 - 16:04:57 -0300)
...
ERROR: cannot open /boot/lua/loader.lua: invalid argument.

Should i just wait for a new recovery firmware to be released or what is the guidance?
I'm just starting out learning pfSense so it's not a big deal to wait a little while...

Thanks,
Stevie.

KOM

@stevie_0815 Install with UFS. When the fix comes out, you can backup your config, reinstall with ZFS, then restore your config. Boom, done.